When a burst of packets enters the network, the existing capacity of the network may not be sufficient to support the traffic which leads to congestion in the network. The packet loss is one of the main problems during transmission which affects the performance of the system. If congestion is detected in advance, the packet loss can be avoided by reducing the packet generation rate at source with effective measures. The existing protocols are predefined mapping between the observed state and the corresponding action. When there is a packet drop in the network (observed state), the congestion window is reduced (action) irrespective of other parameters related to the networking environment such as resource utilization by each user, moving average, etc. Therefore, these protocols are unable to adapt their behaviour in the new environment or learn from past experience for better performance. To overcome these issues, the Machine Learning (ML) technique is required in the field of networking to learn from past experience and analyze the current network scenario to take certain actions. ML has the ability to deal with huge amounts of complex data which becomes one of the reasons for applying ML in the field of networking.